230 lines
8.2 KiB
Python
230 lines
8.2 KiB
Python
# Copyright 2025 ByteDance and/or its affiliates.
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
|
|
|
|
from typing import Any, Optional, Tuple
|
|
|
|
import torch
|
|
import torch.nn.functional as F
|
|
from torch import nn
|
|
|
|
|
|
def precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):
|
|
freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))
|
|
t = torch.arange(end, device=freqs.device) # type: ignore
|
|
freqs = torch.outer(t, freqs).float() # type: ignore
|
|
freqs_cis = torch.polar(torch.ones_like(freqs), freqs) # complex64
|
|
return freqs_cis
|
|
|
|
|
|
def reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):
|
|
ndim = x.ndim
|
|
assert 0 <= 1 < ndim
|
|
assert freqs_cis.shape == (x.shape[1], x.shape[-1])
|
|
shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]
|
|
return freqs_cis.view(*shape)
|
|
|
|
|
|
def apply_rotary_emb(
|
|
xq: torch.Tensor,
|
|
xk: torch.Tensor,
|
|
freqs_cis: torch.Tensor,
|
|
) -> Tuple[torch.Tensor, torch.Tensor]:
|
|
xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))
|
|
xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))
|
|
freqs_cis = reshape_for_broadcast(freqs_cis, xq_)
|
|
xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)
|
|
xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)
|
|
return xq_out.type_as(xq), xk_out.type_as(xk)
|
|
|
|
|
|
class AdaLNZero(nn.Module):
|
|
def __init__(self, dim):
|
|
super().__init__()
|
|
self.silu = nn.SiLU()
|
|
self.linear = nn.Linear(dim, dim * 6)
|
|
self.norm = nn.LayerNorm(dim, elementwise_affine=False, eps=1e-6)
|
|
|
|
def forward(self, x, emb=None):
|
|
emb = self.linear(self.silu(emb))
|
|
shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = torch.chunk(emb, 6, dim=1)
|
|
x = self.norm(x) * (1 + scale_msa[:, None]) + shift_msa[:, None]
|
|
return x, gate_msa, shift_mlp, scale_mlp, gate_mlp
|
|
|
|
|
|
class AdaLNZero_Out(nn.Module):
|
|
def __init__(self, dim):
|
|
super().__init__()
|
|
self.silu = nn.SiLU()
|
|
self.linear = nn.Linear(dim, dim * 2)
|
|
self.norm = nn.LayerNorm(dim, elementwise_affine=False, eps=1e-6)
|
|
|
|
def forward(self, x, emb):
|
|
emb = self.linear(self.silu(emb))
|
|
scale, shift = torch.chunk(emb, 2, dim=1)
|
|
x = self.norm(x) * (1 + scale)[:, None, :] + shift[:, None, :]
|
|
return x
|
|
|
|
|
|
class Attention(nn.Module):
|
|
def __init__(self, encoder_dim, encoder_n_heads, max_seq_len):
|
|
super().__init__()
|
|
self.encoder_n_kv_heads = encoder_n_heads
|
|
model_parallel_size = 1
|
|
self.n_local_heads = encoder_n_heads // model_parallel_size
|
|
self.n_local_kv_heads = self.encoder_n_kv_heads // model_parallel_size
|
|
self.n_rep = self.n_local_heads // self.n_local_kv_heads
|
|
self.head_dim = encoder_dim // encoder_n_heads
|
|
|
|
self.wq = nn.Linear(
|
|
encoder_dim,
|
|
encoder_n_heads * self.head_dim,
|
|
)
|
|
self.wk = nn.Linear(
|
|
encoder_dim,
|
|
self.encoder_n_kv_heads * self.head_dim,
|
|
)
|
|
self.wv = nn.Linear(
|
|
encoder_dim,
|
|
self.encoder_n_kv_heads * self.head_dim,
|
|
)
|
|
self.wo = nn.Linear(
|
|
encoder_n_heads * self.head_dim,
|
|
encoder_dim,
|
|
)
|
|
|
|
def forward(
|
|
self,
|
|
x: torch.Tensor,
|
|
start_pos: int,
|
|
freqs_cis: torch.Tensor,
|
|
mask: Optional[torch.Tensor],
|
|
):
|
|
bsz, seqlen, _ = x.shape
|
|
xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)
|
|
xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)
|
|
xk = xk.view(bsz, seqlen, self.n_local_kv_heads, self.head_dim)
|
|
xv = xv.view(bsz, seqlen, self.n_local_kv_heads, self.head_dim)
|
|
|
|
xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)
|
|
xq = xq.transpose(1, 2) # (bs, n_local_heads, seqlen, head_dim)
|
|
keys = xk.transpose(1, 2) # (bs, n_local_heads, cache_len + seqlen, head_dim)
|
|
values = xv.transpose(1, 2) # (bs, n_local_heads, cache_len + seqlen, head_dim)
|
|
|
|
output = F.scaled_dot_product_attention(xq, keys, values, mask[:, None, None, :], is_causal=False)
|
|
output = output.transpose(1, 2).contiguous().view(bsz, seqlen, -1)
|
|
return self.wo(output)
|
|
|
|
|
|
class FeedForward(nn.Module):
|
|
def __init__(
|
|
self,
|
|
dim: int,
|
|
hidden_dim: int,
|
|
multiple_of: int,
|
|
ffn_dim_multiplier: Optional[float],
|
|
):
|
|
super().__init__()
|
|
if ffn_dim_multiplier is not None:
|
|
hidden_dim = int(ffn_dim_multiplier * hidden_dim)
|
|
hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)
|
|
|
|
self.w1 = nn.Linear(
|
|
dim, hidden_dim
|
|
)
|
|
self.w2 = nn.Linear(
|
|
hidden_dim, dim
|
|
)
|
|
|
|
def forward(self, x):
|
|
return self.w2(F.silu(self.w1(x)))
|
|
|
|
|
|
class TransformerBlock(nn.Module):
|
|
def __init__(self, encoder_dim, encoder_n_heads, max_seq_len):
|
|
super().__init__()
|
|
self.encoder_n_heads = encoder_n_heads
|
|
self.encoder_dim = encoder_dim
|
|
self.head_dim = encoder_dim // encoder_n_heads
|
|
self.attention = Attention(encoder_dim, encoder_n_heads, max_seq_len)
|
|
self.feed_forward = FeedForward(
|
|
dim=encoder_dim,
|
|
hidden_dim=2 * encoder_dim,
|
|
multiple_of=256,
|
|
ffn_dim_multiplier=None,
|
|
)
|
|
self.attention_norm = AdaLNZero(encoder_dim)
|
|
self.ffn_norm = nn.LayerNorm(encoder_dim, elementwise_affine=False, eps=1e-6)
|
|
|
|
def forward(
|
|
self,
|
|
x: torch.Tensor,
|
|
t: torch.Tensor,
|
|
start_pos: int,
|
|
freqs_cis: torch.Tensor,
|
|
mask: Optional[torch.Tensor],
|
|
):
|
|
"""
|
|
Perform a forward pass through the TransformerBlock.
|
|
|
|
Args:
|
|
x (torch.Tensor): Input tensor.
|
|
start_pos (int): Starting position for attention caching.
|
|
freqs_cis (torch.Tensor): Precomputed cosine and sine frequencies.
|
|
mask (torch.Tensor, optional): Masking tensor for attention. Defaults to None.
|
|
|
|
Returns:
|
|
torch.Tensor: Output tensor after applying attention and feedforward layers.
|
|
|
|
"""
|
|
# pre-norm & modulation for attention input
|
|
norm, gate_msa, shift_mlp, scale_mlp, gate_mlp = self.attention_norm(x, emb=t)
|
|
|
|
# attention
|
|
attn_output = self.attention(norm, start_pos, freqs_cis, mask=mask)
|
|
|
|
# process attention output for input x
|
|
h = x + gate_msa.unsqueeze(1) * attn_output
|
|
|
|
norm = self.ffn_norm(h) * (1 + scale_mlp[:, None]) + shift_mlp[:, None]
|
|
ff_output = self.feed_forward(norm)
|
|
out = h + gate_mlp.unsqueeze(1) * ff_output
|
|
|
|
return out
|
|
|
|
|
|
class Transformer(nn.Module):
|
|
def __init__(self, encoder_n_layers, encoder_dim, encoder_n_heads, max_seq_len):
|
|
super().__init__()
|
|
# Decoder
|
|
self.layers = torch.nn.ModuleList()
|
|
for _ in range(encoder_n_layers):
|
|
self.layers.append(TransformerBlock(encoder_dim, encoder_n_heads, max_seq_len))
|
|
|
|
self.norm = AdaLNZero_Out(encoder_dim)
|
|
self.out_proj = nn.Linear(encoder_dim, encoder_dim)
|
|
|
|
# Rope embedding
|
|
freqs_cis = precompute_freqs_cis(
|
|
encoder_dim // encoder_n_heads, max_seq_len
|
|
)
|
|
self.register_buffer("freqs_cis", torch.view_as_real(freqs_cis), persistent=False)
|
|
|
|
def forward(self, x, t, attn_mask, start_pos=0):
|
|
freqs_cis = torch.view_as_complex(self.freqs_cis.float())[start_pos: start_pos + x.size(1)]
|
|
for i, layer in enumerate(self.layers):
|
|
x = layer(x, t, start_pos, freqs_cis, attn_mask)
|
|
x = self.norm(x, t)
|
|
x = self.out_proj(x)
|
|
return x |